--;.

,
162
PREDICTION OF FORTEAN EVENT REPORTS
FROM POPULATION AND EARTHQUAKE NUMBERS
by Michael A. Persinger, Ph. D.
INTRODUCTION
R
EPORTS of odd events, generally classified under the
vague label of Fortean Phenomena, remain a chal
lenge to science. Are they consequences of: (1) artifacts
of human language, (2) the normal memory modifications
for odd events (of which most people are not familiar),
(3) poor observation, (4) the statistical refuse of otherwise
orderly natural principles, (5) routine principles with ani-
sotropic geometries of application, or (6) the sporadic
occurrence of phenomena that are symptomatic of other
principles not yet measured or understood?
One or more of these solutions have beeR discussed by
various writers with different levels of analytical prowess.
The majority of pro- and con-Fortean enthusiasts have
used logical arguments or suave semantics in order to
demonstrate their pOint. Since the language used is pri-
marily a nominal scale, composed of 0,1 conditions
(something is either true or not true, without intermediate
conditions), solutions appear as simple and clear demon
strations.
Unfortunately qualitative conclusions from clever argu-
ments can elicit an illusionary sensation of resolution.
One walks away with a simple explanation and with the
anxiety of uncertainty removed, only to be frustrated by
the next barrage of contradictory and frustrating facts.
After a few trials, one begins to suspect the limits of simple
language.
Quantitative analyses, whereby numbers rather than
words are used, contain almost unlimited degrees of free-
dom. Whereas with a scale like: true or not true, one
would be forced to partial all of the different and unrelated
events in the universe into two over-inclusive categories,
numerical scales allow the precise (within measurement
limits) discrimination of the potentially unlimited events.
Using numerical scales, one can seriously predict and
understand complex phenomena.
Quantitative analyses require the experimenter to be
more than just a verbal acrobat. Clear understanding of
methodology becomes more important than mind-blOwing
metaphors, such as "magnetic windows" or "elastic time."
One is faced with a different set of questions that often
alters the complexion of the phenomena.
The present study is the first systematic statistical evalua-
tion of Fortean events as a numerical population. Three
specific questions were considered as basic themes: (l) How
much does population contribute to the numbers of For-
tean reports, (2) How much does earthquake history
contribute to Fortean reports, and (3) Do different Fortean
categories relate in different quantitative ways to these
variables?
The first question is critical since human beings are the
primary observers (measurement devices) and reporters
PURSUIT Fall 1979
of odd events. If one can accommodate the majority of
the variability in odd-event numbers by population alone,
then one is not left with much "numerical room" in which
to test various theoretical assumptions. Statements like
population contributes a little "or a lot" are not sufficient
here. We must know how much of the phenomena is due
to population alone.
The second question is a test of a general theory de-
veloped earlier.! In fact, if one conducts a content analy
sis of Charles Fort's writing,2 the semantic connection
between earthquakes and odd events is conspicuous.
The magnitude of the contribution of earthquake numbers
to odd events (if at all) is critical information to test any
mechanism involved.
The quantitative solution of the third question is essen-
tial for both the methodology and mentality of odd events.
If all categories of odd events are predictable in the same
manner from known variables and display the same numer-
ical characteristics. then the possibility of any differen-
tiation between possible mechanisms is reduced. As a
singular inseparable source of variability, the entire popu-
lation of Fortean events is reduced to an amorphous
mass-smarting significantly of measurement artifact or
random error.
The validity of Fortean events cannot be demonstrated
with the data sets now available to the researcher. One
can only determine reliabilities within limits. In the present
analyses, the events, like any probationary phenomena,
are considered as verbal reports. They should be considered
populations of numbers. Neither the acceptance nor the
rejection' of the phenomena can be based upon a single
case or group of cases.
In an area plagued by measurement problems and by
the human emotions of theoretical commitment, phe-
nomena can be rejected completely or accepted without
question. Some events may be accepted as statistically
significant when they are chance, while other events may
be rejected as chance when they are reliable. Long and
systematic evaluation decreases this
METHODS AND RESULTS
Data Sources
A total of 1,153 odd events reported between the years
1850 and 1973 within the 31 states bordering and east of
the Mississippi river were used as data. These events
were sorted from a larger file of more than 6,000 events
that had been the source for Space-time Transients and
Unusual Events. 1 Approximately 97 % of the 385 events
reported between 1850 and 1945 were taken directly
from the works of Charles Fort.
2
Approximately 98 % of
the 768 events reported between 1945 and 1974 were
collected from Fate magazine. The remaining data were
collected from various historical publications.
These 31 states were selected since their population
values have changed systematically over the years with
the exception of Florida. States to the west are character-
ized by longer histories of few human observers and by
larger areas. Data from Fort and Fate were selected in order
to compare the reliability of odd event types. Analyses
on two separate populations anow internal checks for:
(1) the consistency of any effect isolated and (2) the re-
liability of classification categories, i.e., whether events
recorded as "luminous lights" in the Fort data are similar
to the events recorded as "luminous lights" in the Fate data.
To reduce sampling or biasing errors, all of the events
reported (even those with missing values) in the Fort and
Fate data sources were used. They had been classified
and placed on computer cards according to time (hour,
day, month, year), space (city/town, state/prOVince,
country and continent), category codes, a short verbal
description and source. The use of all events reported in
the sources reduces exclusion/inclusion artifacts from ad
hoc or a priori conclusions about "which phenomena are
real or not real." Obviously, this does not reduce the initial
selection bias of either Charles Fort or of the editors of
Fate magazine. -
Since unusual events tend to cluster within one-month
periods but display inter-event-intervals of up to several
years, simple addition of all events was considered mis-
leading. For example, UFO (unidentified flying object)
reports may be reported in several places within a state
several times during one week. To obviate this potential
artifactual inflation of cases within categories and states,
all events from a given category within a state occurring
within one month were given no more than a value of 1
(no matter how many redundancies).
Although summary lists for each state were available
from information collected for Space-time Transients and
Unusual Euents. a new printout was produced. Events
were counted by this experimenter, using the one-month
criteria for selection. Only general UFO reports (lights in
the sky). feline-like animals and odd glass events (phan-
toms, snipers, Window crosses), displayed any difference
from the original data compilation. This difference was
minimal. Correlations (for states) between the present
and previous analyses were greater than 0.96.
Categories
The data had been coded according to 9 major categories
each containing 9 subcategories. The 9 categories were
unusual: Falls (FALT). sounds (SOUN), electrical events
(ELEC). UFO-related phenomena (UFOT), human occur-
rences (HUMA), animal sightings (ANm, "forces" (TELE) ,
geological displays (GEOP), and archeological finds
(ARCH). The rationale has been discussed elsewhere.
l
For this analyses, 4 to 5 or all subcategories of each
category were combined. The advantage of this oper-
ation is the increased number of events within each cate-
gory. Many of the subcategories contained 0 or 1 event
for more than 25% of the 31 states. Although such nominal
data can be accommodated with some statistical methods.
a single report (even by a group) is suspect when pattern
analyses are involved. Larger sample sizes reduce the
anisotropic contribution of an aberrant case to the popu-
lation. assuming the population approaches some regular
function.
163
The disadvantage of collapsing subcategOries involves
the arbitrary pooling of possibly different phenomena
and mechanisms (even though they had been classified
under the same gross category). If the 9 subcategories
reflected 9 mutually exclusive mechanisms with different
numerical properties, then any unique aspect of each
subcategory could be masked. The problem is similar to
assuming that all phenomena labeled as UFOs are from
a Singular source merely because they share the same
crude (and over-inclusive) verballabel.
3
.
4
The categories or clusters of subcategories used for this
analyses (and sample event types for each) were:
FALA: reported falls of rocks, ice, and unusual rains
from the sky or inferred falls from immediate proximal
observation.
FALB: reported falls of organic materials, including
live or dead animals (predominately reptiles/amphibians).
FAL T: the combination of FALA and FALB.
ELEe: aurora-like events, mirages of cities and ghost-
lights or related events (luminosities occurring in the same
area, Le., 100 m) over at least one decade but not asso-
ciated with classic haunts.
SOUN: odd sounds: two major sources; first, involving
groans, bumps, and related low frequency sounds; sec-
ond, involving voices with no apparent origins but not
related to haunt situations.
UFOA: lights in the sky, including the reports of an odd
comet or an odd meteor from Fort's rubric.
UFOB: more complicated and discriminative luminous
displays, including phantom, angels, people in the sky
and UFOnauts."
UFOT: the combination of UFOA and UFOB.
HUMA: odd human reports, centered around human
lOCi, such as stigmata, spontaneous human combustion.
ANIA: odd or unusual creatures including feline-like
humanoid, reptilian (sea monster) and flying (large birds)
types.
ANIB: odd behavior of animals or appearance of
known animals such as massive movements, sudden
appearances.
ANIT: the combination of ANIA and ANIB.
TELE: odd force-related events, including phantom
sniper events, bleeding- statues, odd glcissetchings follow:.
ing a discrete event (e.g., a storm). ,
GEOL unusual geological reports, svch as sudden
appearance of rocks in fields, large holes>-(foHowing un-
known explosions) in fields, moving rocks, and alleged
gravity-magnetic anomalies.
ARCH: unusual, unclassified or noncommensurate
(with current theory) archeological findings such as 12
feet tall skeletons, ancient European coins, odd potsherds
of "non-existent" cultures. For more details, see (1). All
consequent statistical analyses were conducted on a
DECSYSTEM 20 computer using SPSS software; sample
results were checked manually.
Comparison of Populations
The means and standard errors of the .mean (S.E.M.)
of each category for the Fort and Fate populations are
presented in Table 1. For comparison, means and S.E.M.s
for the percent contributions of each category to the total
number of events for each population are shown as well.
PURSUIT Fail 1979
Except for the F AL T categories in
the Fort data and the ANIT and UFOT
164
TABLE 1
categories in the Fate data, the percent
distribution of the other categories is
relatively similar for both populations.
In fact, the distortion appears to be
primarily a consequence of the heavy
Means (Xl and standard errors of the mean (S.E.M.) for the members of odd
events in different categories and for the relative contribution (in percent) of each
category to the total population. for both Fate and Fort data (n = 31 states) .
loading from the general UFO (UFOA)
category in the Fate data. If one ex-
cludes this category only from both
populations of data, the discrepancies
are reduced almost to zero except for
the FAL T and AN IT reports.
The correlation coefficients (Pear-
son's r) between categories within Fort
and Fate data are shown in Table 2
while coefficients between Fort and
Fate populations for the different cat- .
egories are shown in Table 3. Scatter-
grams were printed for all correlations
and inspected Visually for non-linear
relationships and extreme values.
Since 4 of the states displayed values
that were beyond the major cluster,
coefficients were calculated for all of
the states and 27 states (without the
4 larger values) .
The n = 31 and n = 27 analyses
FA LA
FALS
FALT
ELEC
SOUN
UFOA
UFOS
UFOT
HUMA
ANIA
ANtS
AN1T
TELE
GEOP
ARCH
ALL
were used throughout the study in order to reduce the
possible artifacts or distortions from one or two extreme
values. Correlation coefficients were calculated for the total
population and for the remaining population after extreme
(large) values had been deleted follOWing visual inspection
of scattergrams. Such operations are important for any
analyses since only 1 value that is 10 times greater on both
the Y and X values in a random cluster, can produce
correlation coefficients of 0.5.
Statistical significance levels are given for different cor-
relation coefficients for the different methods used. Here
p < . a 1 is taken as a criterion that the relationship is not
Raw Data Percent of Total
FORT FATE FORT FATE
1.7 0.3 2.2 0.4 18.9 3.9 8.6 1.3
1.10.3 0.5 0.2 9.2 3.0 1.8 1.2
2.8 0.4 2.7 0.5 28.1 4.5 10.4 1.4
0.60.1 1.0 0.2 5.1 1.4 4.2 1.0
0.3 0.1 0.4 0.1 2.0 0.8 2.0 0.9
1.5 0.3 7.9 1.2 10.7 1.6 29.9 2.5
0.9 0.2 1.4 0.2 8.5 2.6 7.0 1.2
2.4 0.5 9.4 1.3 19.2 2.7 36.9 2.8
2.1 0.7 2.1 0.4 14.2 3.4 8.5 1.6
0.9 0.2 2.3 0.5 6.0 1.2 9.3 1.6
0.3 0.2 4.3 0.7 1.8 1.2 5.2 1.8
1.2 0.3 3.6 :t 0.7 7.8 1.3 14.5 2.0
1.8 0.3 3.4 0.5 13.9 2.3 14.4 2.1
0.4 0.1 1.3 0.2 4.2 1.8 5.0 1.0
0.7 0.2 0.9 0.2 5.6 1.8 4.1 0.9
12.4 2.0 24.8 3.5
a routine chance occurrence. A correlation coefficient
that is significant beyond the .01 level would occur by
chance alone 1 out of 100 times: or, one would expect
that 1 out of every 100 correlation coefficients to be sig-
nificant by chance alone at this level.
. The magnitude of the correlation coefficient required
for statistical significance varies with sample size. For 31
cases (states), a correlation value must be 0.46 or greater
to be significant at the .01 level. For 27 states, the cor-
relation coefficient must be 0.49 or greater. (Both values
assume a two-tailed test considering the nature of the
data.)
TABLE 2
Correlation coefficients (r) between categories in the Fort population (above the diagonal) and in the Fate population
(below the diagonal) . ...",'
FALT
ELEC
SOUN
UFOS
HUMA
ANIT
TELE
GEOP
ARCH
FALT ELEC SOUN UFOS HUMA ANIT TELE GEOP ARCH
0.08 0.45 0.48 0.22 0.45 0.02 0.24
0 . .c;v--_1...: .. 00""-_.Q.06 0.46 0.30 0.34 0.13 0.21 0.28
0.12 0.25 0.14 - 0.18 - 0.15 0.08
0.30 0.22 - 0'1J3"-_J..0tJ--J).07 0.04 0.09 - 0.02 0.19
0.50 0.36 0.19 01"0"'--1JXr--.Q . ..9 0.34 0.14 0.50
0.60 0.41 0.17 0.30 0.b'i---.J.0(J""--!l..54 0.38 0.54-
0.60 0.36 0.09 0.33 0.65- OAS"'---l.OO--_Q,12 0.29
0.43 0.37 0.43 0.58 0.40 0.22 0]'5"'--l.:OO---QJ1
0.54 - - 0.02 - 0.05 0.43 0.26 0.25 0.30 O.2'g---l..OCJ
-
r 0.49. p 0.1. two-tailed
PURSUIT Fall 1979
One should realize that practical significance and statistical
significance are two separate issues. A correlation coeffi-
cient that is significant statistically may in actual fact "ex-
plain" or account for only a trivial amount of the total
variability in the data. For example, a r of 0.50 between
variable X and Y appears relatively impressive. However,
such a relationship accounts for only 25% (r
2
= (0.50)2)
of the variability of the data.
In other words, one can explain or account for only
25% of the variability in X by knowing Y or vice versa.
The remaining 75% of the variability in the data is asso-
ciated with some unspecified source. Even a correlation
of 0.90 between X and Y indicates that 20 % of the vari-
ability cannot be accounted for by the other variable.
Correlations Among Event Categories,
Population and Earthquakes
Previous analyses of the data indicated that the three
variables: odd events, population and earthquake history
of the 31 states were inter-correlated. The relationship
between population numbers and event numbers is ex-
pected since human beings are the primary measure. 5
The relationship between population numbers and
earthquake numbers is more obscure. However, recent
unpublished data indicate a possible explanation for this
correlation. States with higher seismic history, within the
population analyzed. tend to cluster along the "valley"
regions and river basins of the primary mountain ranges
and have displayed large areas of fertile regions. Not sur-
prisingly, these areas were settled first and have expanded
proportionally.
Based upon previous data. one can offer a probationary
explanation for these interrelationships. Whereas popula-
tion is correlated with event numbers due to the require-
ment of human observers, earthquakes and populations
are interrelated due to a shared third variable: some geo-
logical-geographical condition. No doubt the relationship,
if any. between odd events and seismic activity will be
confounded pervasively by this third variable.
For the present analyses, odd event categories from
the Fort data were correlated with population values for
the year 1900 (the median value for the Fort printout)
and with earthquake numbers between 1700 and 1945
(FQTOT). The Fate data were correlated with population
values for the year 1960 and for earthquake numbers
between 1945 and 1974. The earthquake data were
collected from Coffman and von Hake
6
while the popula-
tion data were taken from U.S. vital statistics.
In order to check the regularity of the correlations be-
tween population numbers for different years, correlations
were completed between population values for every 10-
year interval between 1880 and 1960. All years after
1900 were intercorrelated by more than 0.96. Only 1880
and 1890 showed correlations below 0.90 with the later
periods.
Tables 4 and 5 show the correlation coefficients (Pear-
son's r) between: (1) total numbers of earthquakes and of
odd events in different categories and (2) population
numbers and odd event numbers, for the Fate and Fort
data using the total population (n = 31) and the limited
population (n = 27). As expected. most of the categories
for both populations are correlated significantly with pop-
165
TABLE 3
Correlation coefficients (r) for different categories
between the Fort and Fate data for the different cat-
egories and subcategories. Coefficients for all 31
states and without the extreme states (n = 27)
are shown.
n = 31 n = 27
FALA 0.16 0.18
FALB 0.41 0.35
FALT 0.49" 0.46
ELEC 0.57" 0.44
SOUN 0.04 -0.02
UFOA 0.42 0.13
UFOB 0.18 0.00
UFOT 0.36 0.10
HUMA 0.59" 0.42
ANIA 0.22 0.22
ANtS 0.14 0.00
ANIT 0.22 0.16
TELE 0.56" 0.40
GEOP 0.32 0.21
ARCH 0.26 0.27
POP 0.93" 0.72"
QUAKES L 0.56 0.47
QUAKESH 0.61" 0.57"
p < .01
ulation number. The perSistence of these relationships in
both n = 31 and n = 27 data indicate these correlations
are not due to a few values. This statement is verified by
inspection of scattergrams. .
The primary difference between the Fort and Fate
data is the number of significant correlations between odd
events and seismic activity. None of the event categories
within the Fate data are correlated significantly with the
earthquake numbers during that period. On the other
hand, depending upon the n = 31 or n = 27 population,
several event categOries from tne Foit'data are correlated
with the earthquake activity.
If indeed earthquakes are correlated to some or all odd
event categories, then some "period-deperfdence" should
be evident. In other words, odd events during a particular
interval should be most correlated with earthquake num-
bers during that period. This statement assumes. of course.
that long-term lags-in the order of decades-do not
exist between the odd events and quake numbers.
As a check for this possibility, total numbers of earth-
quakes recorded before 1874, between 1875 and 1909
and between 1910 and 1945 were correlated with the dif-
ferent categories in the Fort data; these three intervals
were called QT1874, QTl909 and QT1945, respectively.
The greatest changes for the three intervals, respectively.
were for FALT (0.46, 0.44, and 0.26), UFOT (0.57,
0.62. and 0.13). HUM A (0.72, 0.36, and O.02). and
ARCH (0.62. 0.46, and 0.28).
The correlation coefficients between the individual
intervals displayed a large range. Quake numbers for the
PURSUIT Fall 1979
TABLE 4
166
variability is. a central operation, sep-
arate factor analyses were completed
Correlation coefficients (r) between different event categories and population (in
the year 1900) and total earthquake numbers (1700 to 1945) for the data collected
from Fort for both n = 31 and n = 27 state analyses.
on the n = 31 and n = 27 population
for each population of data. Extreme
values have been known to grossly
distort factor isolations.
n = 31 n = 27
The main 9 categories were factor
analyzed. Major analyses included:
P1900 FQTOT Pl900 FQTOT ELEC, FALT, SOUN, UFOT,
HUMA, ANIT, TELE, GEOP, ARCH.
In most instances, total categories
were substituted by one of their major
divisions. UFOT was replaced by
UFOB, FALTwas replaced by FALA
and ANIT was replaced by ANIA.
These replacements involved the
"most strange" of these categories.
For a given category (variable) to be
significantly loaded upon a factor,
a correlation coefficient of greater
than 0.4 on that factor was required.
FALA 0.64* 0.38 0.52* 0.04
FALB 0.39 0.55" 0.44 0.68*
FALT 0.71" 0.61"
0.63* 0.50
ELEC 0.34 0.42 0.23 0.32
SOUN 0.07 0.17 -0.10 0.09
UFOA 0.56* 0.57 0.42 0.38
UFOB 0.23 0.53" 0.15 0.47
UFOT 0.51" 0.60* 0.38 0.56"
HUMA 0.70* 0.55 0.53" 0.48
ANIA 0.74 0.43 0.48 0.31
ANIB
0.37 0.53" 0.09 0.25
ANlT 0.66 0.51 0.39 0.35
TELE 0.64 0.48" 0.49" 0.31
Results of the factor analyses for
the Fort and Fate data are shown in
Table 6. The loading of each variable
on the factors are shown in paren-
theses. Both eigen values and the
percent of the total variability of all
GEOP 0.28 0.08 0.40 0.03
ARCH 0.54 0.61" 0.33 0.59"
P1900:FQTOT = 0.51 PI900:FQTOT = 0.35
QT1874 interval correlated 0.60 with total quake numbers
and less than 0.40 with any of the other intervals. The
other three intervals correlated between 0.75 and 0.83
with total quake numbers. Correlation coefficients between
the low (Mercalll N and V) and total quakes for the QT1874,
QT1909, QT1945, and QT1973 intervals were 0.91,
0.83,0.94, and 0.98, respectively. On the bases of these
data, one would expect little difference between total and
IV -V magnitude correlations with odd event categories.
Factor Analysis
The perSistent correlation between the different odd-
event categories and population, and for that matter, the
inter-correlation between categories, suggested that factor
analysis was in order. Given an array of correlation co-
effiCients for a set of variables, factor analyses allow the
experimenter to determine whether some underlying
pattern of relationship exists. If this relationship exists,
then the data could be "rearranged" or reduced to a smaller
set of factors.
Although the single most distinctive characteristic of
factor analysis is its data-reduction capability. the isolation
of factors allows potential isolation of the source uariables.
Using factor analysis, one can find groups of variables
that display shared properties of variability. If completed
carefully and systematically, sometimes one can isolate
the mechanism associated with the shared source of
variability ..
Factor analyses involve a large number of different pro-
cedures. In this study, the most accepted method: PA 2
is used.
7
All factor loadings were determined from the
varimax solution, which gives the clearest resolution
(enhances high or low loadings on a factor). Again, since
PURSUIT Fall 1979
the categories accommodated by each
factor are shown in parentheses beside
the factor numbers. Eigen values are measures of relative
importance of the factor. (Eigen values of 1.00 or less
indicate that the factor explains no more than that ex-
p a i n e d by a single variable.)
Two to three factors emerged from the n = 31 and
n = 27 populations for both the Fort and the Fate data.
The cumulative variability that can be accommodated by
these factors ranged from 64% to 73 %. Despite the vari-
able loadings on some factors. some perSistent patterns
are evident. For example, the variables UFOB, FALT
and ELEC appear on the same factors for both Fate and
Fort data and for both n = 27 and n = 31 analyses.
Similarly, some v.ariables, such as TELE, FAL T and UFOB
are loaded on 'more than one factor significantly. AU four
analyses also show the persistence of a cluster composed
ofFALT, HUMA, ANIT, TELE.
To determine how population and earthquake num-
bers might load on these factors, FQTOT and P 1900
were added to the Fort data while QT1973 and P1960
were added to the Fate list. The results were clear and
perSistent for the Fort data. Whereas the TELE. FAL T,
HUMA, ANIT cluster was loaded heavily by population
(0.72) and lightly by earthquakes (0.41), the factor con-
taining UFOB, FALT and ELEC was loaded heavily by
earthquakes (0.70) but not by population (0.38). The
third factor was loaded by earthquakes (0.45) but only
with the n = 31 analyses. Substitution of QT1909 (1875-
1909) did not alter the contribution of the quake variable.
The Fate data were different with respect to population
and to quake loadings. For these data, population was
loaded heavily on Factors 1 (0.70) and 3 (0.40) for both
analyses and also on Factor 2 for the n = 27 analyses.
Earthquakes during the Fate data period were loaded in
a negative way on Factor 3 only for
the n = 27 (- .5) and n = 31 (- .60)
167
TABLES
analyses. The addition of the popula-
tion and quake data produced a third
factor for the n = 31 analyses only.
Interestingly, Factor 3 in the Fate data
contains very high positive loadings
of the UFOS and ELEC categories.
The addition of UFOA did not alter
the factor pattern for either data pools.
Correlation coefficients (r) between different event categories and populations
(1960) and total earthquake numbers (1945-1974) for the data collected from Fate
magazine for both n = 31 and n = 27 state analyses.
To determine whether some of the
categories from both the Fort and
Fate data may fall on the same factors,
all major categories of the two data
pools were factor analyzed together.
The existence of 18 variables rather
than 9 variables is expected to inflate
the number of factors discriminated.
However, with careful interpretation,
artifacts can be reduced.
The results of these analyses are
shown in Table 7. The six factors that
emerged can account for 75 % to
78% of the variability. Again, the first
factor tended to be weighted by the
HUMA, TELE, FALT, and ANIA
categories (Is indicate Fort data, 2s
indicate Fate data). Note that Factor 4
again emerged with ELECl, UFOBl,
and FALT l.
FALA
FALB
FALT
ELEC
SOUN
UFOA
UFOS
UFOT
HUMA
ANIA
ANIB
ANIT -
TELE
GEOP
ARCH
p < .01
The difference between the n = 27 and n = 31 analy-
ses are quite evident as well. The former results show
little sharing of categories from both Fort and Fate data
on the same factors. A given factor is predominated by
either Is (Fort) or 2s (Fate). However, the n = 31 analyses
show a clear clustering effect whereby both Is and 2s
appear on a single factor such as Factor 1.
As a preliminary identification of these factors, earth-
quake histories for the Fort and for the Fate data and
populations for the two periods were added to the analy-
ses. Only the first factor was loaded by population: P1900
(0.83) and P1960 (0.77). Fortean period earthquakes
were loaded slightly on Factor 1 (0.44). Factor 4. was
loaded by Fortean period earthquakes (0.67) but not by
population. Since P1900 and P1960 were so highly inter-
correlated, the same analyses were conducted with only
one populapon value (PI900 or PI960); the basic pattern
did not change. The clustering of Fortean period quakes
on UFOSl, FALTI and ELECI remained persistent.
On the basis of these results, some new variables were
computed for further analyses in order to determine the
contribution of population and earthquake history (if any)
to their occurrence. Categories that appeared on the
same factors were computed (added) and given new
labels. These computed categories (for Fort and Fate
separately) were:
PIEZ = (UFOB + FALT + ELEC)
LUMO = (UFOS + ELEC)
MIND = (HUMA + TELE)
EART = (FALA + ELEC + SOUN + GEOP)
FORC = (FALA + TEL E)
ODDS = (UFOT + ANIA + ARCH)
n = 31 n = 27
P1960 QT973 P1960 QT973
0.69" 0.27 0.61" 0.19
0.44 0.26 0.39 0.17
0.70" 0.31 0.63" 0.21
0.66" 0.01 0.54" -0.30
0.36 0.25 0.38 0.20
0.81" 0.11 0.76" -0.11
0.64" 0.02 0.10 0.13
0.82" 0.10 0.74" -0.15
0.67" 0.22 0.62" 0.06
0.48" 0.10 0.47 0.03
0.60" -0.03 0.48 -0.24
0.65" 0.06 0.59" -0.12
0.71" 0.28 0.65" 0.11
0.59" 0.23 0.59" 0.16
0.42 0.14 0.32 0.15
P1960:QT973 = 0.30
P1960:QT973 = 0.01
EVER = (ALL - UFOA)
ALL = (All 9 total categories)
The inter-correlations for some of these between Fort and
Fate data are shown in Table 8.
These computations have both positive and negative
aspects. One could argue that the different variables (cat-
egories) by themselves do not reflect an "integrated"
phenomenon, whatever it is, since they are artificial frag-
mentations due to labeling systems. The computation
presents the data as an integrated pool. On the other
hand, one could argue that adding similar source data
merely amplified any effect artifactually. Such limits should
be realized.
Partial Correlation
Inspections of the data demonstrated th;'e association of
population numbers, earthquake numbers, odd event
numbers, and, implicitly total areas (per state). Partial
correlation provides the researcher with a single measure
of association describing the relationship between two
variables while adjusting for the effects (on variability)
from one or more other variables. 7
Partial correlation can be a helpful tool to unmask spur-
ious correlations. A spurious correlation is defined as a
relationship between two variables X and Y in which
X's correlation with Y is purely a consequence of X vary-
ing with some other variable Z, where Z is the true predictor
of Y. If the effects of Z were controlled (partialled out),
the X would no longer vary with Y.
To use an example critical of Fortean phenomena in
general, assume the relationship between odd events and
earthquakes is due to the contribution of population to
PURSUIT Fal11979
168
odd events. If the contribution of population was held
constant, then the correlation between earthquakes and
odd events would no longer be statistically significant or
as large. In this situation, the original correlation between
odd events and earthquakes would be considered spurious.
the n = 31 data, respectively. The results of correlating
PIEZ1 with quakes and controlling for population, on the
other hand, produced correlations of 0.49 and 0.55, re-
spectively. One can conclude that the controlling variables
for PIEZ1 in this situation would be primarily earthquake
numbers during the Fortean period and secondarily pop-
ulation numbers.
Table 9 shows the first order partial correlation coeffi-
cients for the Fort and Fate data according to events,
population and quake numbers, for different computed
categories. The results of using all states (n = 31) and
states within the major cluster (as inspected by scatter-
gram) are shown as well. Partials for both Fortean period
(QFORT) quakes (1874 to 1945) and Fate period (QFATE) .
quakes (1945 to 1974) are shown to compare populations.
Taking PIEZ1 for example, the partial correlations with
population numbers but controlling for quake numbers
show a correlation of 0.38 and 0.40 for the n = 27 and
PIEZ1, LUM01, 00051, ALL01 and EVERl, all dis-
played some correlation with earthquakes even when
controlled for population. This was evident to some extent
when the data from the Fort period were correlated with
earthquakes for the Fate period. On the other hand,
none of the Fate period computed categories were sig-
nificantly correlated with quakes from either period.
Population was the primary control variable.
These data support in general the results of the factor
TABLE 6
Major factors and. relative loadings of categories (in parentheses) for n = 31 and n = 27 state analyses for both Fate and Fort data.
Factor 1 (3.14.35%)
FALT .(0.62)
ELEC (0.67)
UFOS (0.71)
Factor 1 (3.66.41 %)
FALT (0.48)
HUMA (0.88)
ANIT (0.96)
TELE (0.47)
ARCH (0.42)
Factor 1 (4.26,47%)
FALT (0.44)
ElEC (0.53)
UFOB (0.63)
HUMA (0.72)
ANIT (0.78)
TElE (0.55)
Factor 1 (4.90.55%)
FALT (0.61)
UFOB (0.69)
HUMA (0.77)
TELE (0.88)
GEOP (0.48)
PURSUIT Fall 1979
FORT
n = 27
Factor 2 (1.44.16%)
FALT (0.48)
ANIT (0.38)
TElE (0.88)
n = 31
Factor 2 (1.13. 14%)
FALT (0.71)
ELEC (0.59)
UFOS (0.42)
TElE (0.42)
FATE
n = 27
Factor 2 (1.26. 14%)
FALT (0.72)
UFOS (0.43)
TELE (0.40)
ARCH (0.71)
n = 31
Factor 2 (1.22.14%)
FALT (0.48)
ElEC (0.72)
UFOS (0.65)
ANIT (0.71)
GEOP (0.47)
ARCH (0.80)
Factor 3 (1.16. 13%)
ANIT (0.83)
GEOP (0.46)
Factor 3 (1.04. 12%)
TElE (0.76)
ARCH (0.41)
Factor 3 (1.04. 12%)
SOUN (0.40)
UFOS (0 .. 40)
TELE (0.40)
GEOP (0.94)
------- -------_ .....__ .. _ .._--
169
TABLE 7
Major factors and relative loadings of categories (in parentheses) for n = 31 and n = 27 analyses for Fort and Fate data
combined (n = IS variables) .
Factor 1 (5.1S, 29%)
ElEC 1 (0.52)
TElE 2 (0.51)
HUMA 2 (0.74)
FAlT 2 (0.65)
ANIT 2 (0.92)
Factor 4 (1.48, S%)
ElEC 1 (0.75)
UFOS 1 (0.71)
FAlT 1 (0.52)
Factor 1 (6.67.37%)
ElEC 1 (0.60)
ElEC 2 (0.64)
UFOS 2 (0.64)
TElE 1 (0.67)
TELE 2 (0.77)
FAlT 1 (0.70)
FAlT 2 (0.73)
ANIT 1 (0.72)
ANIT 2 (0.67)
HUMA 1 (0.65)
HUMA 2 (0.72)
GEOP 2 (0.66)
ARCH 1 (0.49)
ARCH 2 (0.53)
Factor 4 (1.3S, S%)
ELEe 1 (0.46)
UFOS 1 (0.52)
n = 27
Factor 2 (2.49.13%)
TElE 1 (0.67)
HUMA 1 (0.60)
ARCH 1 (0.57)
ANIT 1 (0.84)
Factor 5 (1.34, 7%)
ElEC 2 (- 0.44)
GEOP 1 (-0.50)
ARCH 2 ( 0.67)
FAlT 1 ( 0.44)
n = 31
Factor 2 (2.37. 13%)
ELEC 1 ( 0.40)
ElEC 2 ( 0.46)
UFOS 2 (- 0.44)
HUMA1 (-0.64)
ARCH 2 ( 0.5S)
ANIT 1 (-0.55)
ANIT 2 ( 0.50)
Factor 5 (LIS. 7%)
UFOS 1 (0.51)
Factor 3 (1.93. 11 %)
SOUN 1 (-O.4S)
UFOS 2 ( 0.79)
HUMA1 (-0.49)
GEOP 2 ( 0.54)
Factor 6 (1.21. 7%)
SOUN 2 (0.69)
GEOP 2 (0.60)
Factor 3 (1.43. 8%)
UFOS 2 (-0.43)
GEOP 1 ( 0.70)
ARCH 1 ( 0.53)
Factor 6 (l.12. 6%)
SOUN 2 ( 0.41)
UFOS 2 - 0 . 4 9 ~
analyses. Some of the categories of the Fort data are
clearly correlated with earthquake and population histories
of the states. Other categories are almost totally loaded
on the population factor, an observation that suggests
a different source for some of the odd events. However,
all the categories from the Fate data appear to be con-
trolled (predicted) by only population data.
The areas of the states are not randomly distributed.
Consequently, the possibility exists that some artifact of
area could be contributing in a negative or a positive
manner to the patterns. Dividing the different event cate-
gory, population and earthquake numbers by total area
of each state. reduces these values to a common unit of
area. In addition, this operation introduces a qualitatively
different type of relationship: density.
Although a common operation by many scientists, the
division by a common factor can be a source of spurious
correlation too. If three uncorrelated populations of data
Xl, X2 and X3 are involved and a person divided Xl by
X2 to obtain 11 and divided X3 by X2 to obtain 12, then
the correlation between II and 12 can approach 0.5.
8
By dividing through by a common source, even a ran-
dom population, the total amount of random distribution
is reduced. For populations that are inter-correlated, the
solution becomes less determinant. i.e .. the division may
enhance or cancel the effect.
If completed appropriately. this procedure can en-
hance the relationship and remove confounding factors
that could be masking any effect. Consequently. first
order partial correlations were calculated for events/ total
PURSUIT Fal11979
170
TABLES
Correlations between Fort and Fate data as a function. of different data transfor-
mations for sample computed categories (following factor analyses), earthquake
and population numbers.
Total Numbers Events/ Area Events/Population
Event n=31 n=27 n=31 n=27 n=31 n=26
PIEZ 0.41 0.23 0.48* 0.39 0.20 0.23
EART 0.48" 0.56* 0.63* 0.23 0.42 0.44
FORC 0.60 0.44 0.66 0.63' 0.11 0.30
MIND 0.67* 0.39" 0.79* 0.68' 0.38 0.50*
ODDS 0.44 0.25 0.50* 0.38 -0.11 0.30
ALLO 0.61* 0.41 0.73* 0.53* 0.46' 0.48
QUAKE 0.61* 0.57' 0.71* 0.47 0.69* 0.41
POP 0.93" 0.72' 0.96* 0.94
TABLE 9
area (of each state) as a function of
population/total area and earth-
quakes/area. and (2) events/pop-
ulation as a function of earthquakes/
population and area/population, for
each of the computed categories.
Except for the significant contribution
of seismic activity to the FORel
(0.58), the basic pattern following
division by area did not change from
Table 9.
Not surprisingly, the division by
population drastically altered the pat-
tern of significant first order partial
correlations. Significant partials for
quakes (controlling for area/popuia-
tion) appeared for PIEZI (O. 75). EART1 (0.58), EVERI
and ALLOI (0.66) and for FORC2 (0.50) and for ALL02
and EVER2 (0.62). All other correlations were not signifi-
cant and predominately negative. Similar patterns held
for both n = 31 and n = 27 analyses.
Multiple Regression
Ultimately, we would like to predict the various cate-
gories of unusual events from known and measurable
variables. Sometimes simple correlations between two
variables do not involve all of the necessary variance
required to explain or to predict one of the variables. In
these instances, one can use multiple regression whereby
more than one variable is used to predict a given depen-
dent variable. 7
Multiple regression allows the calculation of a multiple
r value that is the consequence of combining the different
variables in the equation. Like the simple r, r2 is the amount
of variability in the dependent variabte that can be accom-
modated by the others. For example, if the r2 for Y is
0.80 with two variables X and Z in the equation, then
(0.80)2 or 64% of the variance in Y can be explained by
the combination of X and Z.
The relative contribution of each variable in the equa-
tion to the prediction of Y is also calculable. Sometimes
all of the major variability can be accounted for by one of
the two (or more) variables. Other times. depending upon
the phenomenon, each variable may add similar amounts
of predictability to the total. Although separately, the two
or more variables in the equations may only marginally
correlate with Y, together, they may produce substantial
multiple correlations.
A primary limit of multiple regression occurs when two
or more variables in the equation are themselves highly
inter-correlated. "Highly inter-correlated" is an arbitrary
term but is usually applied to r values greater than 0.90.
In this situation, the high inter-correlation between X and
Z themselves, may artifactually inflate the multiple r value
forY.
Using multiple correlation, we can predict how much
of the variability in the odd events can be accounted for
by the two variables in question: population and seismic
activity. If one can obtain multiple rs of 0.90 or greater
then one can predict safely that 81 % of the variance in
the odd events can be predicted by knowing population
and earthquake measures. Only 20% of the variability in
odd events would be due to some other source/sources.
With other data in the social sciences, this variability 'is
related usually to measurement error or scaling "noise."
Table 10 displays multiple rs for sample Fort and sample
Fate populations. The two most populated single cate-
gories: UFOT and HUMA are presented for comparison.
As can be seen, the multiple rs range from 0.64 to 0.87
for the Fort data and from 0.67 to 0.87 for the Fate data.
Translated into predictability, however, one can predict
only between 40% to 76% of the variability in different
categories as a consequence of knowing population and
earthquake history.
The two populations: Fort and Fate. differ blatantly
with respect to the relative contribution of earthquake
numbers and population to the total variability of the cat-
egories. In some of the Fort categories, earthquakes contrib-
171
TABLE 10
Multiple regreSSion results demonstrating multiple r values
and the relative contribution (r
2
) of either population or
earthquake numbers during the appropriate periods for
different categories in the Fort and Fate data.
FORT FATE
FortQ FateQ FortQ Fate Q
PIEZ
MR 0.80 0.68 0.84 0.83
Qr
2
53% 30% 9% 0%
P r2
10% 16% 62% 68%
LUMO
MR 0.60 0.34 0.81 0.80
Qr2
35% 4% 2% 5%
P r2
1% 8% 64% 59%
FORe
MR 0.74 0.73 0.81 0.81
Qr
2
24% 53% 16% 11%
P r2
30% 0% 51% 56%
UFOT
MR 0.70 0.56 0.83 0.84
Qr2
45% 16% 5% 2%
P r2
4% 15% 68% 68%
HUMA
MR 0.74 0.70 0.67 0.67
Qr2
30% 20% 1% 1%
P r2
24% 32% 45% 45%
EVER
MR 0.87 0.83 0.86 0.87
Qr2
51% 26% 0% 0%
Pr
2
26% 43% 75% 75%
ute, as shown by their r2 values, most of the explained
variance. Popula!ion appears trivial. Again, in the Fate
data, all of the variance is due to population.
From the point of view of prediction, (even with pos-
sible spurious correlations not totally eXftuded), the mul-
tiple regreSSions completed with events/area as a function
of population/area and earthquakes/area show much
greater potency. As seen in Table 11, a multiple r of 0.98
is found with the Fate data. In other words, 96 % of the
variability for all basic categories (with the exception of
UFOA) when combined can be predicted from population/
area and earthquake/ area. Whereas population by itself
has an r of 0.93 with the EVER category. the earthquakes
add an extra 0.05. This is the difference between ex-
plaining 86% of the variance and 96% of the variance.
In order to enhance the prediction of the Fort data in
particular, since the multiple rs were smaller than the Fate
categories, various substitutions were made in population
and quake variables. Since the Fort data were scattered
over a longer period of time than the Fate data, one might
argue that the use of population for 1900 or the large
PURSUIT Fall 1979
r

172
TABLE 11
Multiple rs and relative I..ontribution of either population or earthquake numbers for sample computed event cat-
egories for both n = 31 and n = 27 analyses. F refers to F -values.
FORT FATE
n = 31 n = 27 n = 31 n = 27
rv F rv
EVER/A
MR 0.83 31 0.83
Q/A r2 62% 11 63%
PIA r2 7% 7 6%
PIEZ/A
MR
0.78 .
18 0.77
Q/A r2
57 23 58%
PIA r2 3% 2 2%
LUMO/A
MR 0.48 4 0.48
Q/A r2 16% 8 16%
PIA r:l
7% 2 7%
FORC/A
MR 0.52 5 0.53
Q/A r2 13% 1 13%
PIA r2 15% 6 15%
MIDWD/A
MR 0.91 71 0.93
Q/A r 46% 1 48%
PIA r2 39% 63 38%
quake interval FQTOT would not display sufficient res-
olution. However, the use of population figures for the
years 1880, 1890, or 1910 did not alter the multiple rs.
The specific use of QT1909 also did not alter the predic-
tion potency. .
Neither the use of VI or more quakes. quakes of IV or
V, total epicenters, or grand total quake numbers en-
hanced the Fort or Fate predictions. In most cases, the
multiple rs were less. The use of these values in the Fate
data reduced the high multiple rs to within the 0.7 to 0.8
range. Reduction of data to events/population as a function
of quakes/population and area/population did not allow
greater multiple r values.
CONCLUSION AND DISCUSSION
Several repeatable patterns have emerged in the dif-
ferent Fortean event categories analyzed in this study.
Simple correlations, partial correlations. factor analyses
and multiple regressions all demonstrate the inter-correla-
tion of Fortean phenomena with population. Although all
these methods do share similar assumptions of internal
PURSUIT Fall 1979
F rv F rv F
27 0.98 295 0.98 305
8 9% 56 9% 54
5 86% 112 88% 108
21 0.95 118 0.96 154
27 13% 34 13% 43
2 76% 198 80% 257
4 0.40 3 0.38 2
7 6% 2 10% 1
2 10% 5 5% 4
5 0.90 60 0.91 57
1 43% 1 44% 1
5 37% 57 39% 54
75 0.93 83 0.93 77
1 8% 15 6% 11
7 78% 33 81% 32
operation, the replicable nature of the data patterns in-
creases the general application of the results.
The interrelati90ship between different odd event cat-
egories within either the Fort or Fate data as well as be-
tween the Fort and Fate data are relatively small. Not
more than 50% (average 15%) of the' yariability in any
category within the same data pool ca\' be predicted or
accounted for by another category in the same pool.
Similarly, a given category in the Fort data can account
for not more than 50% of the variability in the Fate data.
The average explained variability between the Fort and
Fate data was about 20 %
Since the population numbers in the Fort data can
account for up to 86 % of the variability of population
numbers in the Fate data (or vice versa), one might con-
clude that the low inter-correlations between odd event
categories imply a minimum contribution from population.
However, this conclusion is inappropriate for a number
of reasons.
First. the variability around the means of each category
for each state in the Fort or the Fate data is more com-
mensurate for the population data. Whereas the popu-
lation measures actually involve millions of counts (and
the entire population, statistically speaking), the odd
event numbers for each state are only samples. Presumably,
all the odd events, by nature of their space-time transient
characteristics, have not been counted.
Since the magnitude of the correlation and the amount
of variability accounted for by a predictor are a function of
the shared patterns of variability, less random variation
occurs in the population data due to the sample size.
When the population data are correlated with the odd
events, the extra sources of random fluctuation from the
odd event sampling, decreases the shared variability
patterns.
Second, the more complicated analyses all demonstrate
that population can account for almost all the variability
contributing to the correlations with earthquakes in most
categories, especially for the Fate data. Since one is more
likely to conclude that population numbers are predictors
of odd events rather than the odd events are predictors of
numbers of people, the population factor appears to
account for most of the variability in all odd event cate-
gories. Even in categories (within the Fort data) where
earthquakes can account for some of the variabil1ty, pop-
ulation is still a contributor.
Certainly, one expects population to correlate with the
numbers of events since the human being is the primary
measurement. It is now clear, for the Fate data at least,
that the relative sums of all Fate events can be accounted
for almost entirely by population variability. When con-
verted to events/area and population/area, in fact, almost
all the variability in the odd events can be accounted for
by population. Without this transformation, more than
75% of the variability is explained.
For the Fate data, at least, one can see that population
is the control variable. One would be arguing on very
precarious ground if explanations of Fortean events were
couched in some environmental or theoretical variable.
Such a variable would have to almost perfectly match the
variability patterns of population to remain hidden.
Statistically, we <.:annot proceed any further on the
population question. The reduction of Fortean phenomena
to purely population causes-in a mechanistic sense-
does not follow totally although it is implied. At present,
we cannot resolve whether the data imply perSistent arti-
facts in the human population or whether odd events are
occurring almost homogeneously over that space and the
population merely reflects the statistical observation.
The differential contribution of earthquake numbers to
the categories within the Fort data strongly suggest that
some of the categories may involve different sources of
variability (and perhaps mechanism). Specifically, PIEZ
and ODDS (and EVER, since it contains these subsets)
showed very clear association with the earthquake history
of the Fort period.
The association with earthquake numbers is not arti-
factual to the period, since even the Fate quake numbers
correlated with the Fortean categories. Any contribution
was expected to be lower, (it involves a different time
period) but still evident (since Fort and Fate period quakes
are correlated). In short, one could conclude that the
composite of UFOB, FALT, ELEC, and ARCH are in-
fluenced by earthquake history.
A more conservative interpretation would be that the
173
earthquake contribution is trivial and perhaps artifactual.
Suspiciously, similar categories for both Fate and Fort
data show similar multiple rs when both quakes and pop-
ulation are in the equation, even though they may con-
tribute to various degrees. One must ask why should the
variability explained by quakes in the Fortean data be
replaced by a population component in the Fate data?
This pattern could be explained as a consequence of
just sampling problems. With large samples (and Fate
contained twice as many cases as Fort data) ,. the errors of
sampling are decreased. As one approaches the entire
population and the number of cases increase, the dis-
crepancy of sampling error is decreased. One would
expect, even predict, the obvious. The EVER category
should show the greatest prediction and correlation with
population since it contains the greatest number of samples.
In short, variability in population can account for most
of the distribution in the different odd-event categories.
No doubt, one could argue that the composite nature of
the analyses removed individual cases and potential
phenomena. However, the smaller the sample size, espe-
cially in this area, the more one is dependent upon the
consequences of random variation, error and worst of all:
anecdotal reasoning.
At a probabilistiC level, one can consider the human
being or the entire population of human beings as a series
of responses. Statistically, one expects reports of some-
thing odd within and throughout the population. Verifica-
tion attempts, from photographs to multiple reporting by
"reliable witnesses." does not alter the mechanisms of
statistical occurrence.
The occurrence of these events would be random;
their pursual would be comparable to playing dice. Every
so often, the enthusiast would "guess" the next number.
Although due to chance, the potent reinforcement would
maintain the fragments of some theory. With each' roll or
each event, the theory, by necessity, must become more
and more far-fetched. Except for word games, there is no
pattern in randomicity.
Although the analyses by spatial (states) increments
were confounded by population variables, the isolation of
some environmental contributor to Fortean phenomena
may still be possible. The lise of .temporal increments
would allow the inclusion of variability associated with
dynamic processes. Time lagged correlations for Fortean
categories could discriminate any sequerttiai relationships
in the data. This procedure has been used successfully
with recent UFO reports.
9
A solution to the Fortean problem, like other dilemmas
in the history of science, will require repeated numerical
analyses and the generation of specific hypotheses
5
that
can be tested with numbers. 'Time-pumps," "magnetic
Windows," and "Supersargasso Seas" are hypnogogic
diversions. They do not solve the problem.
REFERENCES
1. Michael A. Persinger and Gyslaine F. Lafreniere, Space-
time Transients and Unusual Events (Chicago: Nelson-Hall,
1977).
2. Charles Fort. The Complete Books of Charles Fort (New
York: Dover, 1974).
PURSUIT Fall 1979
3. M. A. Persinger, Limitations of human verbal behavior in
context of UFO related stimuli, in R. F. Haines (Ed.). UFO
Phenomena and the Behavioral Scientist (Metuchen. New
Jersey: Scarecrow Press. 1979).
4. M. A. Persinger, The problems of human verbal behavior:
the final reference for measuring ostensible psi phenomena. The
Journal of Research in Psi Phenomena, 1976, 1. 72-79.
5. M. A. Persinger, Possible geophysical sources of close UFO
encounters: expected physical and behavioral-biological effects.
In R. F. Haines (Ed.)' UFO Phenomena and the Behavioral
Scientist (Metuchen. New Jersey: Scarecrow, 1979).
6. J. L. Coffman and C. A. von Hake (Eds.), Earthquake
AN APPENDIX TO
174
History of the United States. (Publication 41-1; Revised Edition
thoroughly 1970. Government Printing Office, Washington.
1973).
7. Norman H. Nie. C. H. Hull, J. G. Jenkins, K. Steinbrenner
and D. H. Bent, SPSS (Statistical Package for the Social Sci-
ences). (New York: Mc-Graw-Hill. 1975, 2nd edition).
8. G. U. Yule and M. G. Kendall, The Theory of Statistics.
(London: Charles Griffin and Co., 1965, 14th edition), pp.
330-331.
9. M. A. Persinger, Earthquake activity and antecedent UFO
report numbers lin submission).
"THE ONE PHYSICAL EXPERIMENT SCIENCE
CANNOT EXPLAIN"
(Printed in Pursuit, Vol. 12, No.3. Summer 1979)
by T. B. Pawlicki
A
MONG the innumerable speculators who have
formulated a holographic model to explain
universal mechanics, Tom Bearden is best known
to me for analyzing the holographic structure math-
ematically. Our theories are virtually identical except
that Bearden has calculated an abrupt, right-angle,
orthorotation at the point where radiant energy is
transformed into a standing-wave vortex; Bearden
calculates one orthorotation on each of the three
axes of field spin; an orthorotation on one axis trans-
forms radiant energy into the mental signals of
words and music; orthorotation on two axes trans-
forms radiant energy into visual, photonic images;
the structure of the spherical standing-wave becomes
manifest with orthorotation on all three field axes.
In contrast, my models demonstrate a gradual rota-
tion of field energy into' the standing-wave phase
and back to the radiant phases, although
is stopped on each of the three axes to bring words,
music and mental images into manifestation. It is
difficult to gainsay a sound mathematical calculation,
but it is perhaps more difficult to deny the validity of
constructed models. Paradoxes of this kind usually
provoke critics into assuming that one or both theo-
ries must be wrong because both cannot be right.
But paradoxes of this kind, including the wave/part-
icle dichotomy, arise simply because mutually ex-
clusive mechanics are both right. The problem is to
perceive the larger perspective in which both views
become synthetic instead of contrary. If you take
the time to study waves upon the surface of a sizable
body of water, you will see a natural physical model
demonstrating the mechanics described by both
Tom Bearden and myself. As a tent is an involute
of a bubble, each wave and wavelet on the surface
of water is an involute of the standing-wave vortices
in a Plate Flutter Experiment tuned to white sound.
The wave always rises at right angles to the veolcity
of radiant energy on its proper scale. The rise is
always abrupt upon encountering an opposing vel-
ocity on the same axis. A wave-front is formed by
opposition on two axes; the third axis cannot be
observed on a plane surface of water. There is the
phenomenon corresponding to Bearden's calculation
of orthorotation, while the standing-wave phase of
the field progresses gradually through.out the full
circle or rotation. ,
(These models have defined the uni'versal holo-
gram almost sufficiently to design an analogue time
transport simple enough for home experimenters to
construct, like the Vortex Drive prototype.)
Do yourself a favor. Member George Eberhart did his co-SITUians
a real favor by compiling a 60-page comprehensive index of Pursuit's
contents, from the inception of our journal through 1978. This index is
available to members only; the nominal price of $1.50 barely covers
reproduction and postage costs. Do yourself a favor and get your very
own copy by sending check or money order for $1.50 to: SITU,
P.O. Box 265, Little Silver, NJ 07739.
PURSUIT Fall 1979